Swarm Intelligence

Insect swarms have showed quite fascinating properties. They are often able to find intelligent solutions to complex problems, such as the architecture of a colony, despite the fact that no member has the necessary skills. Each member follow simple rules, no member individually control the swarm, making it extremely robust, it's because of this simplicity and robustness, that swarms have become an interesting model for solving various problems in computer science. Computer scientists and engineers have been able to transform models of collective behavior of social insects into useful methods for optimization and control. A new field of research has emerged which aims to transform knowledge that ethologists have into artificial techniques of problem solving: It's Swarm Intelligence.

This new metaphor for problem solving has become a very hot topic in recent years, and the number of successful applications is growing exponentially in Combinatorial optimization, communications networks, and in robotics.
Researchers are being increasingly interested in this new way of reaching another form of artificial intelligence, swarm intelligence, the emergence of a collective intelligence, from groups of simple agents.

Definition and Background:

*Swarm: Group of agents in which the exchange of information impacts the individual behavior, allowing for the realization of global objectives out of reach of an agent. E.g. ant colony.

*The term "Swarm Intelligence" was coined by Gerardo Beni in 1989 (Proceedings of the Seventh Annual Meeting of Robotics Society of Japan): 

Swarm Intelligence is a property of systems of non-intelligent robots exhibiting collectively intelligent behavior evident in the ability to unpredictably produce specific (i.e., not in a statistical sense) ordered patterns of matter.

In this definition "robot" denotes an entity capable of presenting a mechanical behavior insomuch as computational, including thus artificial systems inasmuch as natural. The critical term in this definition is the word "unpredictable"; In fact, only when a system is considered "unpredictable" in the production of an order, that means something to speak of as intelligent.

If "unpredictable" is the key element in the definition of the intelligent behavior of the system, the key element in the definition of swarm intelligence in the system is the presence, of a global characteristic, which is absent in the units that form the system.

The expression "Swarm Intelligence" was first used by Beni, Hackwood and Wang, in the their recherches on the simulation of self-organizing agents, and this, in the context of cellular robotic systems. From a historical perspective, the idea of using collections of simple agents or automata to solve optimization and control problems on graphs and computer networks, etc. was alreadyy present in the works of Butrimenko, Tsetlin, Stefanyuk and Rabin. Tsetlin identified the important characteristics of bio-inspired automata which makes the approach based on the swarms of insects potentially powerful, the randomness, decentralization, indirect interactions between agents and elf-organization. Butrimenko applied these ideas in the control of telecommunication networks and Stefanyuk in the cooperation of radio stations.

In 1990, a more extensive definition was  provided by Bonabeau, Dorigo and Theraulaz, they supposed the term should be applied on:

Any attempt to design algorithms or distributed problem-solving devices inspired by the collective behavior of social insect colonies..and ther animal societies.

Agassounon, Martinoli and Easton (2004) suggest that swarm intelligence takes its inspiration: "From biological examples provided by social insects.. such as ants, termites, bees and wasps, and by swarming, flocking, herding and shoaling phenomena in vertebrates."

By contrast, Martinoli (1999) asserts that such bio-inspiration is not necessary and that the defining characteristic of swarm intelligence should be an emphasis on local control and communication (as opposed to global), claiming that:

Swarm intelligence arises from local interactions and is based on local information and communication mechanisms. (Martinoli,1999)

So generally, anything that is inspired from swarms is swarm intelligence. In artificial intelligence, swarm intelligence refers to the collective intelligence of insect societies, if we sum up:

"Swarm intelligence consists of studying and building societies of simple artificial agents that are able to collectively produce a complex response. In such a multi-agent system, each agent has a limited view of the system, but it decides autonomously. Thereby, the system is characterized by a decentralized working: Each agent follow simple rules, no agent decide or coordinate actions of other agents."

***************